This study, which is to our knowledge the first systematic pathologic analysis focusing on this issue, elucidates that, among resected PanNETs, by applying standardized counting methodology Ki67 index ≥ 10% identifies a significantly more aggressive and a metastasis-prone group and as such supports the evolving management protocols in the oncology literature that advocate managing this group differently.3,7,9,24,25,26,27,28,29,30,31,32,33,34,35 This category also stood the multivariate analysis along with lymph node metastasis and tumor size/stage.

It is important to point out that Ki67 index is a continuum; as such, it is plausible to pick essentially any cutoff and possibly find some prognostic value and associations. However, the current cutoff of 3–20%, which had been mostly extrapolated from G1 PanNETs proved to have virtually no discriminatory value, such that G1 and G2 PanNETs are currently mostly regarded together as one category (G1/2) for management purposes, leaving almost 95% of cases unstratified. In this study, 10% cutoff was selected and tested for two main reasons: (1) This cut off has recently begun to be employed in the oncology literature to select patients for therapy by somatostatin receptor analogues and others3,7,9,24,25,26,27,28,29,30,31,32,33,34,35; (2) The analysis of the primary cohort in our study in which Ki67 count was performed with special care revealed that with the “maximally selected rank statistics” method the proper cutoff was 12%. Considering this number is fairly close to the rule-of-thumb number of 10%, this figure was selected and indeed proved to have very striking discriminatory value.

Accordingly, evaluating G2 PanNETs that have a Ki67 index rate ≥ 10% separately as G2b (i.e., cases with Ki67 index of 10% to ≤ 20%), this group was found to exhibit significantly higher rates of metastasis than G2a group (i.e., Ki67 of 3% to < 10%). Other signs of aggressiveness also were significantly higher in the G2b group. Perhaps more importantly, these characteristics of G2b group was very similar to the G3 (>20%) group. Increasingly, the oncology literature encourages the consideration of therapy for any G3 PanNETs.28,45 Considering that the G2b cases were found to be very similar to G3 (both in the initial as well as the validation cohorts in this study), the approach applied to G3 cases may have to be considered for the G2b group, which needs to be further investigated by clinical trials. Clearly, if nothing else, this group is a candidate for closer follow-up. It should be reemphasized that his group is currently hidden in the wide G2 category and typically managed along with the G1 cases under the G1/2 umbrella; thus, their aggressiveness skips clinical attention.

The fact that G2b group is comparable to the G3 group suggests that the prognostic impact of Ki67 index peaks above 10% range but also starts to level off after a certain point probably as it approaches to the 20% range. In fact, in the primary cohort, the signs of aggressiveness were even higher in the G2b group than the G3 group. However, this may be related to the small number of cases in that initial cohort, because this difference mostly disappeared when combined with the validation cohort. Regardless, the findings are strongly indicative that G2b group defined is highly comparable to the G3 cases. It should be noted that, although well-differentiated PanNETs are viewed to be molecularly entirely distinct from PDNECs, recent studies are finding G3 cases to have some molecular overlaps with PDNECs5,46 with 35% of the cases showing p53 alteration. Considering G2b group shows aggressive behavior similar to G3, it will be important to analyze this group at the molecular level to see if it bears cases that show overlaps with (or ability to transition to) the PDNECs.5

It is important to reiterate that, in the current daily practice, the differences in aggressiveness between the G1 and the current broad G2 category (3–20%) is not viewed to be striking enough by the majority of the community to justify a differential treatment based on the grade alone.7,25,47 However, this study demonstrates that, when the characteristics of the G2b group is compared to G1, with the lymph node metastasis rate being several folds higher, and liver-distant metastasis rate of very compelling nine- to tenfold (58.5% vs. 6.1%), the significance of separating the G2b category and distinguishing it from the G1 becomes much more crucial.

These results also reemphasize some of the evolving practice recommendations for pathologists. First, the impression that “there is not much difference between a Ki67 of 4% and 19%” (all currently regarded G2), and “in fact, there is not much difference between G1 and G2 PanNETs” which in turn leads to the common application of eye-balling method in daily practice, is no longer valid. The striking differences in the behavior of G2a versus G2b group (which is even more dramatic when G2b is compared with G1) elucidated in this study necessitates more careful counting and numerical documentation of the Ki67 index. It should be reiterated that, after all, Ki67 index is a continuous variable, and that 9% (regarded G2a in this study) is probably not really that different than 11% (regarded G2b). As in any grading or staging system, there are imperfections in the cutoff regions. For this reason, in our opinion, documentation of the specific count (not only the grade) in the surgical pathology report is warranted. This would allow the management team to assess other findings, for example, the presence of other risk factors such as the highly infiltrative pattern16 or oncocytic phenotype,12 and incorporate age, patient expectation/choices and comorbidities to determine the course of action putting all these factors together. For this reason, we believe it is crucial to provide a specific number for Ki67 index, and we also recommend reporting of infiltration pattern and phenotypic classification, all of which altogether could aid in the final clinical decision making, especially for the cases that stand on the fence. Of note, despite the acknowledged shortcomings in applying cutoffs to a continuous process, at the same time, as in any grading and staging system, a cutoff to stratify the patients also is needed for Ki67. In this study (in which the hard stop cutoff was found to be 12% by the “maximally selected rank statistics”) when combined with the evolving impression in oncology literature regarding 10%, points to this rule of thumb number of 10% as the most practical and valid cutoff.24,25,26,27 As such, all of this warrants the creation of a G2b category.

The findings in this study will have significant impact on the treatment of PanNET patients. For management of tumors that have already been resected, it has not been clear which radiologic test to employ, how soon, and with which frequency. The rates of metastasis elucidated in this study for G2b cases (with three quarters of this group showing lymph node metastasis, and almost two thirds showing liver/distant metastasis, the latter 8 times more frequent than in the G1 cases), clearly indicates that this group warrants close follow up at minimum. In fact, depending on the age of the patient and other factors, the eligibility of G2b cases to a management that is similar to G3 (for which somatostatin analogues and other therapies are increasingly being employed as adjuvant treatment) may have to be considered, although, of course, this requires further studies.24,25,26,27,28 However, with a rate > 60% showing liver/distant metastasis, for this group, it is clear that surveillance protocols need to be modified; at minimum frequent liver radiograms are warranted for the G2b group, probably starting with 3-month intervals and extending them as time goes by, which is what we currently recommend for our patients. In contrast, for a G1 PanNET, which overall has less than 10% liver/distant metastasis rate, the surveillance may not have to be that intense, and an adjuvant therapy would obviously not be justifiable.

The results of this study also has significant implications on the management of PanNETs that are smaller than 2 cm. Currently, watchful waiting is considered as an option of these cases, regardless of the grade.10,11 Indeed, in this study, for < 2 cm tumors that are G1, the liver/distant metastatic rate at the follow up period was < 2%, and for even G2a, it was < 10%. However, when it came to the proposed G2b group, the rate of metastasis was 50% even in these small tumors. Granted this constitutes a relatively small percentage of the cases (7% of < 2 cm PanNETs are G2b and 3% G3), but nevertheless it is not a trivial number. These data indicate that if such a case is encountered, resection should be considered strongly.

Naturally, these approaches will have to be modified as more data comes out for the different subgrades, but in the absence of any specific data at this time, the approaches described above are justifiable based on this data and current literature. Clinical trials are warranted to consolidate new guidelines for these groups.